490

F. GNE, M. A. BELEN, P. MAHOUT, ET AL., SIGNAL AND NOISE MODELLING OF MICROWAVE TRANSISTORS

Signal and Noise Modeling of Microwave Transistors

Using Characteristic Support Vector-based Sparse
Regression
Filiz GNE, Mehmet Ali BELEN, Peyman MAHOUT, Salih DEMREL
Electronic and Communication Engineering Dept., Yldz Technical University, Istanbul, Turkey
gunes@yildiz.edu.tr, mbelen@yildiz.edu.tr, pmahouti@gmail.com, salihd@yildiz.edu.tr
Manuscript received November 4, 2015

Abstract. In this work, an accurate and reliable S- and

Noise (N) - parameter black-box models for a microwave
transistor are constructed based on the sparse regression
using the Support Vector Regression Machine (SVRM) as
a nonlinear extrapolator trained by the data measured at
the typical bias currents belonging to only a single bias
voltage in the middle region of the device operation domain of (VDS/VCE, IDS/IC, f). SVRMs are novel learning
machines combining the convex optimization theory with
the generalization and therefore they guarantee the global
minimum and the sparse solution which can be expressed
as a continuous function of the input variables using
a subset of the training data so called Support Vector
(SV)s. Thus magnitude and phase of each S- or N- parameter are expressed analytically valid in the wide range of
device operation domain in terms of the Characteristic SVs
obtained from the substantially reduced measured data.
The proposed method is implemented successfully to
modeling of the two LNA transistors ATF-551M4 and
VMMK 1225 with their large operation domains and the
comparative error-metric analysis is given in details with
the counterpart method Generalized Regression Neural
Network GRNN. It can be concluded that the
Characteristic Support Vector based-sparse regression is
an accurate and reliable method for the black-box signal
and noise modeling of microwave transistors that
extrapolates a reduced amount of training data consisting
of the S- and N-data measured at the typical bias currents
belonging to only a middle bias voltage in the form of
continuous functions into the wide operation range.

Keywords
Scattering S-parameters, noise N-parameters, Support
Vector Regression Machine (SVRM), Generalized
Regression Neural Network (GRNN)

1. Introduction
Fast and accurate models of microwave devices and

DOI: 10.13164/re.2016.0490

antennas are indispensable in contemporary microwave

engineering. In todays RF and microwave technology,
there is an ever-increasing demand for higher level of system integration that leads to massive computational tasks
during simulation, optimization and statistical analyses,
requiring efficient modeling methods so that the whole
process can achieve the required reliability. However
modeling still remains a major bottleneck for efficient RFMicrowave CAD. Among all component models, unreliable transistor models can easily lead to unsuccessful design
because of their strong influence on the overall circuit
performance. Thus, the efficiency of such models in terms
of accuracy and speediness is critical to assure a reliable
design.
Artificial Neural Network (ANN)s have emerged as
the valuable tools to extend the repertoire of the statistical
methods. Particularly the Back-Propagation Multi-Layer
Perceptron (BPMLP)s have been employed for the nonlinear interpolation based on learning from the measured
or simulated data, in the fast, accurate and reliable modeling of both active and passive microwave devices [16].
Todays fast, accurate and reliable Signal S- and Noise Nblack-box model of a microwave transistor can be achieved
only by a single simple BPMLP with one hidden layer
which is capable of the simultaneous generalization of 12
Scattering S- and Noise N- functions into the entire operation domain of the bias condition VDS/VCE, IDS/IC and the
frequency f for all the configuration types [3]. However,
these so-called Back Propagation Neural Networks BPNNs
suffer from a number of disadvantages. The most important
disadvantages can briefly be summarized as follows:
(i) The variants of the back-propagation can be shown to
converge to the local minima of the error surface; (ii) As
most algorithms use (pseudo) random numbers, the result
of training varies between runs even for identical networks
and training datasets. Therefore BPNNs in modeling need
statistical analysis of the results of a lot of runs.
In recent years, regression with Support Vector Machine (SVM)s and probability based neural networks
named as Generalized Regression Neural Network
(GRNN) remove these handicaps and facilitate to substantially reduce the expensive fine discrete training data with

CIRCUITS

RADIOENGINEERING, VOL. 25, NO. 3, SEPTEMBER 2016

their typical modeling applications to transistor [710],

integrated microstrip and slot antennas in [11], [12],
printed transmission lines [1315] and vertical interconnects in microwave packaging structures [16].
The SVM based on the Structural Risk Minimization
(SRM) principle is one of the most widely used learning
algorithms, and achieves superior generalization performance for both classification and regression problems [16].
The SVM is systematic and properly motivated by the
statistical learning theory [17]. Training of the SVM involves optimization of a convex objective function and
globally minimizes to complete the learning process without suffering anymore from local minima [17], [18]. In
addition, SVM can handle a large input, and can automatically identify a small subset consisting of informative
points, namely Support Vector (SV)s [17]. Thus sparseness
of solutions is obtained that the large amount of data to be
fully characterized by the set of SVs, a subset of the training set. Especially the working principle of SVRM based
on the small sample statistical learning theory is utilized in
[13], [14] to build a knowledge-based SVRM model for the
synthesis of printed transmission lines as fast as the coarse
models and at the same time as accurate as the fine models.
In those works, the SVs are determined using the coarse
data generator which are the empirical synthesis formulae
of the related transmission line and accuracy of the models
is increased by supplying the corresponding fine values of
the SVs from the fine data generator which are the fullwave EM simulators. Thus, amount of the expensive fine
training data is reduced substantially as pointed out in
[1116].
In this paper, we worked out to determine the characteristic SVs for each S- and N- parameter using the
sparse training data, that is subject to be generalized
throughout a wide range of operation domain. Thus a fast,
accurate and reliable black-box S- and N- parameter
models are constructed based on the sparse regression
using SVRM as a nonlinear extrapolator trained by the data
measured at the typical bias currents of the single bias
voltage in the middle region of the device operation
domain of (VDS/VCE, IDS/IC, f). This modeling method is
facilitated by the two features: The first one belongs to the
black-box characterization parameters which depend upon
the bias currents more than the bias voltages; the second is
the superior generalization ability of the SVRM. On the
other hand, the training time and memory of SVM are
expensive and are strongly correlated to the number of
training patterns , as O( 3 ) and O( 2 ), respectively
[1719]. Thus approximately at least 50% of the training
data is reduced that corresponds to 1/8 and 1/4 of the
training time and the memory complexity respectively of
the SVRM as the nonlinear interpolator in the S- and Nmodeling used in [38]. The proposed method is
implemented on the modeling of a low-noise microwave
transistor ATF-551M4 and VMMK-1225 as a case study.
The paper is organized as follows: Fundamentals of
the SVRM are briefly given in the next section. Third

491

section is devoted to the case study where SVRMs are used

as nonlinear extrapolators in modeling of the microwave
transistors together with error-metric as compared to the
counterpart method (GRNN) performance. The paper ends
with the conclusion.

2. Support Vector Regression

Machines
SVRM builds a nonlinear function between a given
input and its corresponding output in the training data. This
continuous nonlinear relation can be used to predict outputs for the given inputs not included in the training data.
The nonlinear function is learned by a linear learning machine in a kernel induced feature space. As in the classification case, the learning algorithm minimizes a convex cost
function and its solution is sparse. In order to explain the
mathematical framework of the SVRMs, let us consider
a training dataset:

D ( xi , yi ), i 1, 2,..... ,

(1)

where xi R n , yi R , SVRM tries to find the mapping

function f( x ) between the input variable vector x and the

desired output variable y, which are operation parameters
of the device (VDS/VCE, IDS/IC, f), and magnitude and phase
of each S- or N- parameter, respectively in our case. In the

SVRM model, the regression function f( x ) is expressed as

follows [1418]:

f ( x ) wT . ( x ) b

(2.1)

where
T
T

x ( x1 , x2 , ........ xn ) ( x ) (1 ( x ), ( x ), .... N ( x )) . (2.2)

2

This step is equivalent to mapping the x input space

into a new space F { ( x ) / x X n } . Thus, f( x ) is

a nonlinear function in the n dimensioned x input space
and linear function in the N dimensioned F-feature space.
Dimension of the feature space N may be smaller than or
equal to the dimension of the input space n. However,
strategy to follow in the selection of the F-feature space is
to convey the essential information of the original data into
its representation in the new space. Thus, a SVRM can be
constructed in two steps: First, a fixed nonlinear mapping

vector ( x ) transforms the data into a feature space F, and
then the linear machine built in this feature space is used to
perform regression on the data. In this manner, we will

refer to the quantities w and b in (1) as the weight vector

and bias:

w T ( w 1 , w 2 , ........ w N ) .

(3)

The regression function in (2.1) is built by determina

tion of w and b by the strategy seeking to optimize the
generalization bounds. This is relied on defining a loss
function that ignored errors that were within a certain distance of the true value. This type of function is referred to

492

F. GNE, M. A. BELEN, P. MAHOUT, ET AL., SIGNAL AND NOISE MODELLING OF MICROWAVE TRANSISTORS

as -insensitive loss function. The use of -insensitive loss

function has the advantage of ensuring the existence of
a global minimum and the optimization of a reliable generalization bound [17], [18]. Linear -insensitive loss function is given as follows [17], [18]:

f ( xi ) yi
0

(4)
L ( x , y , f )

yi f ( xi , yi ) otherwise
where is predefined. Equation (4) defines tube so that if
the predicted value is within the tube the loss is zero, while
if the predicted point is outside the tube, the loss is the
magnitude of the difference between the predicted value
and target with the radius of the tube. In order to minimize the sum of the linear -insensitive losses the function
can be given as [17], [18]:

1 2

w C L (xi , yi , f )
2
i 1

(5)

to control the size of w for a fixed training set. C is

a parameter to measure the trade-off between the complexity and losses. Since the given objective function (5) is
a convex, it has no local minima and it guarantees the
global minimum which is one of the advantages of the
SVRM on the other regression methods, especially neural
networks. Substituting (4) into (5) and introducing two
slack variables i and i , equation (5) is transformed into
the following soft margin primal optimization problem
[17], [18]:

min
{
imize

w , b ,i ,i

1 2
w C (i i )}
2
i 1

(6a)

subject to

( w, ( x ) b) y , 0,
i
i
i i i

y ( w, ( x ) b) i 1, 2,...,
i
i
i

(6b)

The optimization problem in (6a), (6b) can be solved

more easily in its dual formulation. In this stage, a Lagrangian function L is constructed combining the soft margin
primal objective function in (6a) with the constraints in
(6b) and introducing Lagrangian multipliers i , i , i , i as
follows:

1 T
( w . w ) C ( i i )
2
i 1
i 1

i [ y ( wT . ( x ) b ]
i
i
i
i 1

i [( wT . ( x ) b y ]
i
i
i
i 1

(7)

(i i i )
i
i 1

The function L has a saddle point with respect to the

primal w , b , , and dual positive variables i , i , i , i

i

at the optimal solution. By substituting the saddle point

conditions with respect to the primal variables into the
Lagrangian function L, we have the following equivalent
dual space objective function to be maximized [17], [18]:

W ( i , i ) max L ( w, b, i , i , i , i , i , i )
w , b ,i ,i

i 1

i 1

yi (i i ) (i i )

(8a)

1
( i i )( j j ) ( xi ), ( x j ))

2 i , j 1

subject to
0 i , i C,

( ) 0, i 1,2,..., .
i

i 1

(8b)

In (8a, 8b) the dual variables i , i are expressed in terms

of i , i and C through the saddle point conditions.
The corresponding Karush-Kuhn-Tucker
complementarity conditions are [17], [18]:

i ( w, ( xi ) b yi i ) 0 ,

i ( yi w, ( xi ) b i ) 0 ,

(KKT)
(9a)
(9b)

i . i i . i 0, ( i C ).i ( i C ).i 0 . (9c)

We must find Lagrangian multipliers i , i . For
this purpose, firstly the nonzero Lagrangian multipliers are
determined using the KKT conditions and substituted into
(8a) and (8b), then the obtained equation is maximized
with respect to the Lagrangian multipliers i , i in the
feature space F. From (9a) and (9b) of the KKT conditions,
it follows that for only the samples satisfying

f ( xi ) yi , the Lagrangian multipliers may be nonzero,

and for the samples of f ( xi ) yi , the Lagrangian multipliers vanish. Since products of i s and i s are zero in
(9c) that means that at least one of these terms is zero. The

sampling data ( xi , yi ) that comes with the non-vanishing

Lagrangian multipliers are called Support Vector SVs, thus
there can be at most SVs. These SVs may be named as
Characteristic SVs in our case, since they facilitate to
express a transistor characteristic parameter which is either
signal or noise parameter, analytically in a wide range of

the operation domain. Then the mapping function f ( x )

between the input variable space and the desired output
variable can be expressed in terms of the SVs within the

kernel K ( x , xi ) space as follows:

f ( x)

nSV

i 1

yi K ( x , xi ) b

K ( x , xi ) ( x ), ( xi )

(10)

where Radial Basis Function (RBF) is chosen as the kernel

K ( x , xi ) function in the case study and nSV is the number
of the characteristic SVs ( nSV ) and b is determined

RADIOENGINEERING, VOL. 25, NO. 3, SEPTEMBER 2016

493

using KKT conditions (10a) and (10b). Thus we have for

f ( x ) an analytical function mapping the n-dimensioned xinput space and one-dimensioned yi-output space, in terms
of nSV characteristic SVs within the RBF kernel domain as
follows:

f ( x)

nSV

i 1

yi exp{ x xi

/ 2 2 } b

(11)

where nSV is number of the characteristic SVs obtained by

applying the -tube SVRM selection process into the

training data set D ( xi , yi ), i 1, 2,..... where xi R n ,

yi R . In the next section a small-signal transistor will be
characterized as a black-box for use in small-signal ampli
fication. In Sec. 4, as a case study, f ( x ) will be obtained
for magnitude and phase of each Scattering (S-) or Noise
(N-) parameter for typical LNA transistors ATF551M4 and
VMMK1225 using their manufacturers data belonging to
the typical currents of only a single bias voltage in the
middle region of the device operation domain of (VDS, IDS,
f). Furthermore error-metric analysis of both of the two
transistors will also be given as compared to the counterpart method GRNN which has also been worked out for the
transistor modeling by our research group in [810].

3. A Small-Signal Microwave
Transistor
3.1 Black-Box Characterization Parameters
The black-box representation of a small-signal transistor with its terminations and port reflections is shown in
Fig. 1. Here signal and noise performance of a small-signal

transistor are given by scattering S (VDS, IDS, f), noise

N (VDS, IDS, f) functions in the device operation domain of

bias condition (VDS, IDS) and frequency f. Thus the meas
ured S , N data at the discrete frequencies throughout the
operational band at a bias condition (VDS, IDS) can be arranged in a table-form function as follows:

f1
S (1)
N (1)

S (2)
N (2)
f2
.

(12.1)
(i )
(i )
f
S
N
i

(l )
(l )

S
N
fl

where S ( i ) , N ( i ) are the S , N parameter data at sampling

frequency fi which can be expressed in the following form:
t

(i )
(i )
(i )
(i )
(i )
(i )
S ( i ) S11 , S11( i ) , S12 , S12( i ) , S21 , S21
, S22 , S22
,

(12.2)
t
(i )
(i)
i)
N Fmin( i ) , opt , (Opt
, Rn( i ) / 50 , i 1,2,.....

where i is each data sample, is phase value for the selected S/N parameters, is magnitude value for the selected S/N parameter.

In this work, the S , N data defined by (12.1), (12.2)

belonging to the typical currents of only a single bias voltage in the middle region of the device operation domain of
(VDS, IDS, f) is found to be sufficient for training SVRM.

Then, the characterization vectors S (VDS, IDS, f), N (VDS,

IDS, f), at any desired frequency f of any bias condition
(VDS, IDS) can be obtained from the network output by
inputting that (VDS, IDS) and the frequency f. Thus the transistor under consideration is characterized continuously all
throughout the device operation domain using greatly reduced amount of sampling data.

3.2 Performance Measure Functions

Once the characterization S , N functions are determined, the signal and noise performance of that transistor
are known for any input and output (S, L) termination
couple at any operation condition (VDS, IDS, f). There are
four functions measuring the performance of the smallsignal transistor (Fig. 1) [2]:

(1) Transducer power gain GT of an active device is

defined as the ratio of the power delivered to the load PL to
the available maximum power PAVS from the source that is

the function of the (S, L) termination couple and S

parameters as follows:
GT

2
2

(1 L )
PL
2 (1 S )
. (13)
GT ( S , L , S )
S
21
2
2
PAVS
1 S22 L
1 in S

(2), (3) Min and Mout are mismatching functions that

measure net powers entering into the input and load, re
spectively. Their dependences to the device signal S and
the termination (S, L) couple can be given as follows:
M in

1 in S
Pin
M in ( S , L , S )
,
2
2
PAVS
(1 S )(1 in )

(14a)

M out

1 out L
PL
M out ( S , L , S )
2
2
PAVout
(1 L )(1 out )

(14b)

where the input in and output out reflection coefficient are

given in terms of the device signal S and the termination

(S, L) couple:

S S
(15a)
in in ( L , S ) S11 12 21 L ,
1 S11 L

S S
out out ( S , S ) S22 12 21 S .
1 S22 S

(15b)

(4) The noise figure F of an active device is defined

as the ratio of signal-to-noise ratios available at input and

output; the noise vector N describes the dependence of the

494

F. GNE, M. A. BELEN, P. MAHOUT, ET AL., SIGNAL AND NOISE MODELLING OF MICROWAVE TRANSISTORS

VDS/IDS
2V
2.7 V
3V

10 mA

15 mA

Test

Test

20 mA
Test

Training

Training

Training

Test

Test

Test

Tab. 1. Training and test data set for ATF551M4.

Scattering
Parameters
Fig. 1. Black-box representation of a small-signal transistor
and its port impedances.

S11

transistor noise figure F on the input termination (source)

reflection coefficient S. These are linked through the following relationship:

S21

65

72

S12

72

72

S22

64

72
Number of SVs
(Total Training Data:45 )
Mag.
Phase

Noise Parameters

S opt
R
F ( S , N ) Fmin 4 N
Z 0 (1 2 ) 1
S
opt

Number of SVs
(Total Training Data:72 )
Mag.
Phase
31
72

(16)

4. Case Study
In this work, the proposed method is implemented to
modeling of the signal and noise parameters of the two
transistors ATF551M4 and VMMK 1225 and the library
LIBSVM for Support Vector Machines is employed [20],
[21]. ATF551M4 previously has been modeled using
SVRM in the interpolation process in [7]. On the other
hand, VMMK1225 is properly selected since it is a typical
LNA transistor with the given manufacturers data in
a large operation bandwidth of 245 GHz at the bias currents from the lower range up to the upper range of IDS= 5,
10, 15, 20 mA of the bias voltages of VDS= 2, 3, 4 V [22].
Firstly let us consider ATF551M4. The 24 S- data for
ATF551M4 transistor are supplied ranging in the
0.118 GHz at each of the bias (VDS, IDS) condition where
the bias current is given as IDS= 10, 15 and 20 mA of each
bias voltage of VDS= 2, 2.7, 3 V by the manufacturers data
sheets. Nevertheless, 15 N-data are supplied ranging in the
0.510 GHz at each of the same bias (VDS, IDS) conditions.
In Tab. 1, the training and test data sets are given for the
SVRM modeling with respect to the bias condition. The
data belong to 2.7 V DC bias condition are given to SVRM
model for training purpose only, while 2 and 3 V DC bias
data are used after training process for test/validation purpose. Thus by this means, the total measured data provided
by manufacturer datasheets is separated in two different
datasets for training and test process. In Tab. 2 number of
characteristic SVs is given for each SVRM model of the
magnitude and phase of each S- and N- parameter and the
spread parameter of the radial kernel function in (11) is
taken to be equal to 0.1 for the lowest Mean Absolute Error
MAEs using (17) and (18) taken place in Tab. 3, 4. As seen
from the results given in Tab. 3, 4, the overall performance
of SVRM model is higher than its counterpart method
GRNN, particularly in S- parameter domain, where the
variations of parameters are more complex than N- parameters for same DC bias conditions over the operation frequency bandwidth.

Fmin

45

opt
RN/50

42
45

45
-

Tab. 2. Number of the characteristic SVs in modeling of

ATF551M4.

Parameter
S11
S21
S12
S22

Error Metric
Mean Absolute Error
Relative Mean Error
MAE
RME
SVRM
GRNN
SVRM
GRNN
0.009
0.018
0.013
0.007
0.279
0.433
0.088
0.015
0.007
0.005
0.132
0.052
0.056
0.314
0.228
9.441

Tab. 3. Error metric of scattering parameters for SVRM

extrapolation of ATF551M4.

Parameter

Error Metric
Mean Absolute Error
Relative Mean Error
MAE
RME
SVRM
GRNN
SVRM
GRNN

Fmin

0.023

0.016

0.050

0.036

opt

0.006

0.007

0.016

0.018

opt

0.726

0.971

0.019

0.020

opt

0.031

0.067

0.018

0.029

RN/50

0.002

0.001

0.028

0.010

Tab. 4. Error metric of noise parameters

extrapolation of ATF551M4.

for

SVRM

Comparative Error Metrics in Tab. 3 and 4 use the

definitions given by (17) and (18) in calculation of Mean
Absolute Error MAE and Relative Mean Error RME for
both the SVRM and GRNN methods:
MAE
RME

1
N

T P

1
N

i 1

Ti Pi

i 1

Ti

(17)
(18)

where Pi and Ti are the ith predicted and target phasor

value, respectively and N is the sampling number for the
validation of the SVRM or GRNN model.
In Figs. 2, 3, the predicted scattering parameters obtained from the Characteristic SVbased sparse regression

RADIOENGINEERING, VOL. 25, NO. 3, SEPTEMBER 2016

495

are compared on the Smith chart and polar plane with the
target values for bias conditions (2 V,10 mA) and (3 V,
20 mA), respectively. Furthermore in Figs. 4 and 5, predicted results of the typical parameters S11 phase and S21
magnitude are compared with all the test samples belonging to the bias voltages VDS of 2 V and 3 V, respectively.

18GHz

18GHz

0.1GHz

Fig. 5. S21 magnitude throughout the entire test data domain

of ATF551M4.

In Fig. 6, the predicted Noise parameters of

ATF551M4 using the Characteristic SVbased sparse
regression are compared with the target values for bias
condition (2 V, 10 mA).
Fig. 2. S22 and S11 predicted values vs. measurements at the
bias condition (2 V, 10 mA) for ATF551M4.

0.1GHz

0.1GHz

18GHz

18GHz

Fmin dB

Fig. 3. S12 and S21 predicted values vs. measurements at the

bias condition (2 V, 10 mA) for ATF551M4.

Similar to modeling of ATF551M4, the Characteristic

Support Vector-based Sparse Regression is also implemented to the modeling of another LNA transistor
VMMK1225 within a large operation bandwidth of
245 GHz at the bias currents from the lower ranges up to
the upper range of IDS = 5, 10, 15, 20 mA of the bias voltages of VDS = 2, 3, 4 V [22]. 692 data belonging to the
typical bias currents of 5, 10, 15, 20 mA at the middle bias
voltage 3 V are used for the S-parameter training and the
model is tested the rest 1384 data as given in Tab. 5. The
N-parameters of the VMMK1225 are given within the
range of 217 GHz frequency in the manufacturers
datasheet. Similarly the N-parameter model is built by
training with the central 64 data and tested with the rest
128 data as in Tab. 5. Similarly in case of modeling of
ATF551M4, the data belong to 3 V DC bias condition are
given to SVRM model for training purpose only, while 2
and 4 V DC bias voltage values are used after training
process for test/validation purpose. Thus by this mean, the

Fig. 4.

S11 phase throughout the entire test data domain of

ATF551M4.

(a)

496

F. GNE, M. A. BELEN, P. MAHOUT, ET AL., SIGNAL AND NOISE MODELLING OF MICROWAVE TRANSISTORS

VDS/IDS

5 mA

10 mA

15 mA

20 mA

2V

Test

Test

Test

Test

3V

Training

Training

Training

Training

4V

Test

Test

Test

Test

Tab. 5. Training and test data set for VMMK-1225.

Scattering
Parameters
S11
S21
S12
S22
Noise Parameters
(b)

Fmin
opt
RN/50

Number of SVs
(Total Training Data: 692 )
Mag.
Phase
685
692
181
692
676
666
31
692
Number of SVs
(Total Training Data:64 )
Mag.
Phase
23
38
34
64
-

Tab. 6. Number of the characteristic SV in modeling of

VMMK1225.
Error Metric
Parameter

S11
S21
S12
S22

Mean Absolute Error

MAE

Relative Mean Error

RME

SVRM

GRNN

SVRM

GRNN

0.024
0.124
0.014
0.101

0.036
0.495
0.014
0.106

0.03
0.066
0.22
0.22

0.047
0.26
0.22
0.23

Tab. 7. Error metric of scattering parameters for SVRM

extrapolation of VMMK1225.
(c)
Error Metric
Mean Absolute Error
MAE

Relative Mean Error

RME

SVRM

GRNN

SVRM

GRNN

Fmin

0.048

0.104

0.093

0.153

opt

0.039

0.030

0.095

0.076

opt

0.914

2.385

0.012

0.029

opt

0.262

0.011

0.095

0.096

RN/50

0.010

0.104

0.085

0.153

Parameter

Tab. 8. Error metric of noise parameters

extrapolation of VMMK1225.

for

SVRM

(d)
Fig. 6. Extrapolated prediction of SVRM vs. measurements at
the bias condition (2 V, 10 mA) of ATF551M4:
(a) Fmin, (b) opt, (c) opt phase, (d) RN/50.

total measured data provided by manufacturers datasheets

is separated in two different datasets for training and test
process. The reduced number of characteristic support
vector corresponding to each scattering and noise parameter is given in Tab. 6. As seen from the results given in
Tab. 7, 8, similarly to the case of modeling of ATF551M4,

again for VMMK1225 the overall performance of SVRM

modeling method is higher than its counterpart method,
GRNN. Thus, one can infer that S- and N- parameter modeling based on Characteristic Support Vector-based Sparse
Regression, SVRM is a fast and reliable modeling method
requiring substantially reduced measurements and human
effort.
Tables 7 and 8 give the error metrics of S- and Nparameters compared with the values resulted from the

RADIOENGINEERING, VOL. 25, NO. 3, SEPTEMBER 2016

counterpart method GRNN respectively. In Figs. 7, 8, the

predicted scattering parameters obtained from the characteristic SVbased sparse regression are compared on the
Smith chart and polar plane with the target values for bias

497

condition (4 V, 20 mA). As is seen from Fig. 7, 8, the

SVRM based regression model has a very accurate response over the wide frequency bandwidth. Likewise in
Fig. 7. 8, the response of SVRM model for all test data
200
SVRM
Target

150

45GHz

100
50
0

2GHz

-50
-100

45GHz
-150

2GHz
-200

200

400

600

800

1000

1200

1400

Test Data Set Samples

Fig. 7. S11 and S22 predicted values vs. measurement at the

bias condition (4 V, 20 mA) of VMMK1225.

2GHz

Fig. 10. S21 phase throughout the entire test data domain of
VMMK1225.

2GHz

45GHz
45GHz

Fig. 8. S12 and S21 predicted values vs. measurements at the

bias condition (4 V, 20 mA) of VMMK1225.
(a)

Fig. 9. S11 magnitude throughout the entire test data domain

of VMMK1225.

(b)

498

F. GNE, M. A. BELEN, P. MAHOUT, ET AL., SIGNAL AND NOISE MODELLING OF MICROWAVE TRANSISTORS

Characteristic Support Vectors. This sparseness of solution

facilitates a fine and fast sparse regression for suitable
regression problems such as modeling of microwave transistors; at the same time it also provides a significant tool
to characterize the large amount of data. These two features
provide SVRMs distinctive superiorities to the other classification and regression methods, such as commonly used
BPNN methods.

(c)

(d)
Fig. 11. Extrapolated prediction of SVRM vs. measurements at
the bias condition (4 V, 20 mA) of VMMK1225:
(a) Fmin, (b) opt, (c) opt phase, (d) RN/50.

In this work first time in the literature, SVRMs are

employed successfully as nonlinear extrapolators in the
black-box modeling of Scattering and Noise parameters of
a microwave transistor as counterparts to the Generalized
Regression Neural Network GRNNs [810]. Thus this
work can be considered mainly in the following significant
contribution in the transistor modeling: Magnitude and
phase of each characterization S- or N- parameter can be
expressed as a continuous function in the throughout device operation domain of (VDS, IDS, f) using only a subset of
the reduced training data so-called Characteristic SVs. In
the modeling process, the data measured at the typical
currents of only a single bias voltage in the middle region
of the device operation domain of (VDS/VCE, IDS/IC, f) is
found to be sufficient for training SVRM. Furthermore
a detailed error-metric analysis is made between the two
nonlinear extrapolators SVRM and GRNN. From this
comparative error-metric analysis, it can be observed that
SVRM is superior to the counterpart GRNN as the extrapolation performance, furthermore the SVRM extrapolation results in an analytical expression while GRNN predicts on the probability bases. It can be concluded that in
this work a revolutionary approach is put forward in the
transistor modeling since Characteristic Support Vector
based-sparse regression results in an accurate analytical
expression using a great amount reduction of training data
and human effort.

samples given in Tab. 14, 15 fit snugly into targeted values. Furthermore in Figs. 9 and 10, predicted results of the
typical parameters S11 magnitude and S21 phase are compared with all the test samples belonging to the bias voltages VDS of 2 and 4 V, respectively. In Fig. 11, the predicted noise parameters of VMMK1225 using the characteristic SVbased sparse regression are compared with the
target values for bias condition (4 V, 20 mA). Once again,
the obtained results suggest that the sparse modeling of
microwave transistors with SVRM method is an effective
and efficient method for transistors S- and N- parameters.

We would like to express our special thanks of gratitude to the BAP of Yldz Technical University for founding our researches under project number of 2015-04-03DOP02 Metamaterial Inspired High Performance Microwave Circuit Design for Microwave Sensor Applications
and 2014-04-03-DOP03 Design of Microwave Amplifier
with Non-uniform Microstrip Transmission Lines.

5. Conclusion

References

SVRMs are novel linear learning machines in the

kernel-induced feature space that combine the convex
optimization theory with the generalization, thus they guarantee the global minimum in the optimization procedure.
Furthermore the solution has sparseness, in other words the
solution can be expressed as a continuous function in terms
of an informative subset of the training data so called

Acknowledgments

[1] ZHANG, Q. J., GUPTA, K. C. Neural Networks for RF and

Microwave Design. Artech House Publishers, 2000. ISBN-10:
1580531008
[2] GNE, F., GRGEN, F., TORPI, H. Signal noise neural
network model for active microwave devices. IEE Proceedings
Circuits, Devices and Systems, 1996, vol. 143, no. 1, p. 18.
DOI: 10.1049/ip-cds:19960150

RADIOENGINEERING, VOL. 25, NO. 3, SEPTEMBER 2016

[3] GNE, F., TORPI, H., GRGEN, F. Multidimensional signalnoise neural network model. IEE Proceedings Circuits, Devices
and Systems, 1998, vol. 145, no. 2, p. 111117. DOI: 10.1049/ipcds:19981712
[4] GIANNINI, F. LEUZZI, G., ORENGO, G., et al. Small-signal and
large-signal modelling of active devices using CAD-optimized
neural networks. International Journal on RF and Microwave
Computer-Aided Engineering, 2002, vol. 12, no. 1, p. 7178.
DOI: 10.1002/MMCE.10007
[5] MARINKOVIC, Z. D., MARKOVIC, V. V. Temperaturedependent models of low-noise microwave transistors based on
neural networks. International Journal on RF and Microwave
Computer-Aided Engineering, 2005, vol. 15, no. 6, p. 567577.
DOI: 10.1002/mmce.20102
[6] MARINKOVIC, Z., PRONIC-RANCIC, O., MARKOVIC, V.
Small-signal and noise modeling of class of HEMTs using
knowledge-based artificial neural networks. International Journal
on RF and Microwave Computer-Aided Engineering, 2013,
vol. 23, no. 1, p. 3439. DOI: 10.1002/MMCE.20631
[7] GNE, F., TOKAN, N. T., GRGEN, F. Signal - noise support
vector model of a microwave transistor. International Journal on
RF and Microwave Computer-Aided Engineering, 2007, vol. 17,
no. 4, p. 404415. DOI: 10.1002/mmce.20239
[8] GNE, F., ZKAYA, U., ULUSLU, A. Generalized neuralnetwork based efficient noise modeling of microwave transistors.
In International Symposium on Innovations in Intelligent Systems
and Applications (INISTA 2011). Istanbul (Turkey), 2011, p. 212
216. DOI: 10.1109/INISTA.2011.5946067
[9] MAHOUT, P., GNE, F., DEMREL, S., et al. Efficient
scattering parameter modeling of a microwave transistor using
generalized regression neural network. In 20th International
Conference on Microwaves, Radar, and Wireless Communication
(MIKON).
Gdansk
(Poland),
2014,
p.
14.
DOI:
10.1109/Mikon.2014.6899968
[10] GNE, F., MAHOUT, P., DEMREL, S., et al. A costeffective
GRNNbased modeling of microwave transistors with a reduced
number of measurements. International Journal of Numerical
Modelling: Electronic Networks, Devices and Fields, 2015.
Published
online
in
Wiley
Online
Library
(wileyonlinelibrary.com). DOI: 10.1002/jnm.2089
[11] ANGIULLI, G., CACCIOLA, M., VERSACI, M. Microwave
devices and antennas modeling by Support Vector Regression
Machines. IEEE Transactions on Magnetics, 2007, vol. 43, no. 4,
p. 15891592. DOI: 10.1109/TMAG.2007.892480
[12] CHAUHAN, N., MITTAL, A., KARTIKEYAN, M. V. Support
vector driven genetic algorithm for the design of circular polarized
microstrip antenna. International Journal on Infrared and
Millimeter Waves, 2008, vol. 29, no. 6, p. 558569. DOI:
10.1007/s10762-008-9354-9
[13] GNE, F., TOKAN, N. T., GRGEN, F. A consensual modeling
of the expert systems applied to microwave devices. International
Journal on RF and Microwave Computer-Aided Engineering,
2010, vol. 20, no. 4, p. 430440. DOI: 10.1002/mmce.20448
[14] TOKAN, N. T., GNE, F. Knowledge-based support vector
synthesis of the microstrip lines. Progress in Electromagnetics
Research, 2009, vol. 92, p. 6577. DOI: 10.2528/PIER09022704
[15] GNE, F., TOKAN, N. T., GRGEN, F. A Knowledge-based
support vector synthesis of the transmission lines for use in
microwave integrated circuits. Expert Systems with Applications,
2010,
vol.
37,
no.
4,
p.
33023309.
DOI:
10.1016/J.ESWA.2009.10.021
[16] XIA, L., MENG, J. C., XU, R. M., et al. Modeling of 3-D vertical
interconnect using support vector machine regression. IEEE
Microwave and Wireless Components Letters, 2006, vol. 16,
no. 12, p. 639641. DOI: 10.1109/LMWC.2006.885585

499

[17] VAPNIK, V. N. An overview of statistical learning theory. IEEE

Transactions of Neural Networks, 1999, vol. 10, no. 5, p. 988999.
DOI: 10.1109/72.788640
[18] CRISTIANINI, N., SHAWE-TAYLOR, J. An Introduction to
Support Vector Machines (and Other Kernel-Based Learning
Methods). Cambridge University Press, Cambridge, 2000. ISBN:
0521780195
[19] CAMPS-VALLS,
G.,
GOMEZ-CHOVA,
L.,
CALPEMARAVILLA, J., et al. Robust support vector method for
hyperspectral data classification and knowledge discovery. IEEE
Transactions on Geoscience and Remote Sensing, 2004, vol. 42,
no. 7, p. 15301542. DOI: 10.1109/TGRS.2004.827262
[20] KIM, D., CHO, S. Pattern selection for support vector regression
based response modeling. Expert Systems with Applications, 2012,
vol. 39, no. 10, p. 89758985. DOI: 10.1016/J.ESWA.2012.02.026
[21] CHANG, C. C., LIN, C. J. LIBSVM: A library for support vector
machines. ACM Transactions on Intelligent Systems and
Technology, 2011, vol. 2, p. 27:1-27:27. Software available at
http://Www.Csie.Ntu.Edu.Tw/~Cjlin/Libsvm.
[22] VMMK1225 0.5 to 26 GHz recommended as a low noise EPHEMT in a wafer scale package. Datasheet. [Online] Cited 201604-13. Available at: http://www.avagotech.com/docs/AV021082EN.

About the Authors ...

Filiz GNE received her M.Sc. degree in Electronics and
Communication Engineering from the Istanbul Technical
University. She attained her Ph.D. degree in Communication Engineering from the Bradford University in 1979.
She is currently a full professor in Yldz Technical University. Her current research interests are in the areas of
multivariable network theory, device modeling, computer
aided microwave circuit design, monolithic microwave
integrated circuits, and antenna arrays.
Mehmet Ali BELEN received his M.Sc. degree in Electronics and Communication Engineering from the Sleyman Demirel University in 2011. He has been currently in
Ph.D. program of Yldz Technical University. His current
research interests are in the areas of multivariable network
theory, device modeling, computer aided microwave circuit
design, monolithic microwave integrated circuits, and
antenna arrays.
Peyman MAHOUT received his Ph.D. degree in Electronics and Communication Engineering from the Yldz
Technical University in 2016. The main research areas are
optimization of microwave circuits, broadband matching
circuits, device modeling, computer-aided circuit design,
and microwave amplifiers.
Salih DEMREL has received M.Sc. and Ph.D. degrees in
Electronics and Communication Engineering from Yldz
Technical University, Istanbul, Turkey in 2006 and 2009,
respectively. He has been currently working as an associate
professor in the same department. His current research
interests are among of microwave circuits especially optimization of microwave circuits, broadband matching circuits, device modeling, computer-aided circuit design,
microwave amplifiers.